APPENDIX AVAILABLE ON THE HEI WEB SITE
Research Report 178
National Particle Component Toxicity (NPACT) Initiative Report on
Cardiovascular Effects
Sverre Vedal et al.
Section 1: NPACT Epidemiologic Study of Components of Fine Particulate Matter
and Cardiovascular Disease in the MESA and WHI-OS Cohorts
Appendix H. MESA Exposure and Health Analysis: Additional Text,
Tables, and Figures
Note: Appendices that are available only on the Web have been assigned letter identifiers that
differ from the lettering in the original Investigators’ Report. HEI has not changed
the content of these documents, only their identifiers.
Appendix H was originally Appendix G
Correspondence may be addressed to Dr. Sverre Vedal, University of Washington,
Department of Environmental and Occupational Health Sciences, Box 354695, 4225 Roosevelt Way NE, Suite 100,
Seattle, WA 98105-6099; email: [email protected].
Although this document was produced with partial funding by the United States Environmental Protection Agency
under Assistance Award CR–83234701 to the Health Effects Institute, it has not been subjected to the Agency’s peer
and administrative review and therefore may not necessarily reflect the views of the Agency, and no official
endorsement by it should be inferred. For the research funded under the National Particle Component Toxicity
initiative, HEI received additional funds from the American Forest & Paper Association, American Iron and Steel
Institute, American Petroleum Institute, ExxonMobil, and Public Service Electric and Gas. The contents of this
document also have not been reviewed by private party institutions, including those that support the Health Effects
Institute; therefore, it may not reflect the views or policies of these parties, and no endorsement by them should be
inferred.
This document was reviewed by the HEI NPACT Review Panel
but did not undergo the HEI scientific editing and production process.
© 2013 Health Effects Institute, 101 Federal Street, Suite 500, Boston, MA 02110-1817
APPENDIX G: MESA exposure and health analysis: additional text, tables and
figures
•
MESA nearest neighbor and inverse-distance weighting (IDW): exposure
estimation
•
MESA nearest neighbor and IDW: health analysis
•
Cross-sectional effects on CIMT in MESA by proximity to nearest monitor
•
Analyses of SO4, SO2, and NO3
•
•
Cross-validation statistics
•
Cross-sectional and longitudinal analyses adjusting for NO2 and SO2
•
Cross-sectional and longitudinal analyses for sulfate, nitrate, SO2, and NO2
Cross-sectional effects of sulfur on CAC and CIMT with the interaction of traffic
variables
MESA nearest neighbor, IDW and city-wide average: exposure estimation
Monitoring at MESA Air sites began in July, 2005 and ended in August 2009. PM2.5
speciation data were measured from April 2007 through August 2008. Data from the one year
period from May 2, 2007 to April 16, 2008 was used for this analysis. The mean for each
monitoring site over that period was calculated as the 10 percent trimmed mean (i.e., the top and
bottom 5 percent of data were excluded for the calculation of the mean).
Three different secondary approaches were used to estimate MESA Air study participant
exposure to PM2.5 components: (1) the annual average concentration of the two-week
measurements at the monitor nearest to each study participant’s residence (nearest monitor); (2)
inverse distance weighting (IDW) of all annual average monitor concentrations in each city
relative to each subject’s residence; and (3) city-wide average concentrations based on all
monitors within each city. For all three approaches, subjects within each city residing within 100
meters of either an A1 road (primary limited access or interstate highway) or an A2 road
(primary US or state highway, without limited access), or within 50 meters of an A3 road
(secondary state or county highway), were assigned the average PM2.5 and EC concentrations
measured at that city’s MESA Air roadside monitor. The roadside monitors in these cases were
not used in calculating the exposures of subjects not living close to an A1, A2 or A3 road. For
OC, silicon and sulfur, roadside monitors were included in the calculations for all three of the
exposure estimation methods.
The following PM2.5 component concentrations were included in this analysis: elemental
carbon [EC], organic carbon [OC], silicon and sulfur.EC and OC were selected as reflecting
combustion sources, silicon as an indicator of crustal dust and sulfur as an indicator of sulfate, a
secondary aerosol. To obtain information on temporal trends, PM2.5 component data were
obtained from the Health Effects Institutes (HEI) Air Quality Database website
(https://hei.aer.com/login.php) for 2002 and 2007.
208 subjects (3.3 %) lived within 100 meters of an A1 road, 243 (3.9%) with 100 meters of
an A2 road, and 1,459(23.3%) within 50 meters of an A3 road. A total of 1,774 subjects (28.4%)
were therefore classified as living close to a large roadway. Using the criteria for living close to
a large roadway, the following in each city lived close to a large roadway: 18.5% in WinstonSalem, 59.5% in New York, 20.8% in Baltimore, 22.8% in St.Paul, 31.2% in Chicago, and18.1%
in Los Angeles. Among those not living close to a large roadway, median distance to the nearest
MESA Air monitor was 4.1 km (IQR 4.3). Most (90.3%) of the participants resided within 10 km
of an air pollution monitor.
Appendix Table G.1 shows PM2.5 and PM2.5 component annual average concentrations (µg
/m3) by city and monitor. Appendix Table G.2 and Appendix Figure G.1 shows median (IQR)
and mean (SD), respectively, study subject PM2.5 and component concentrations over all six
cities according to the three exposure metrics. Mean PM2.5 concentrations for the six study sites
based on nearest monitor ranged from 16.22µg/m3(Los Angeles) to 10.26µg/m3(St. Paul); EC
ranged from 2.67µg/m3 (New York City) to 0.70µg/m3 (St. Paul); OC ranged from 4.37µg/m3
(Winston-Salem) to 2.74µg/m3 (St. Paul); silicon ranged 0.15µg/m3 (Los Angeles) to 0.08µg/m3
(Baltimore); and sulfur ranged from 1.65µg/m3 (Baltimore) to 0.85µg/m3 (St. Paul). Medians
(IQRs) for IDW and city average are also shown.
Appendix Figure G.2 shows the correspondence between mean PM2.5 and PM2.5 component
concentrations from the CSN for 2002 and 2007 in the MESA cities with available CSN data.
There was generally good correspondence between concentrations over that five year span
except for silicon, which showed a decrease, except in Los Angeles and Winston-Salem.
MESA nearest neighbor, IDW and city average: health effects analysis
Statistical Analysis:
Multiple linear regression was used to estimate the associations between PM2.5 measures and
CIMT and CAC (among persons with Agatston scores greater than zero). Agatston scores were
analyzed after log transformation. Binomial regression was used to estimate the associations
between PM2.5 measures and the presence of CAC (Agatston Scores>0). All measures of
association were expressed per inter-quartile range (IQR) of each concentration measure.
Covariates in the regression analyses were selected a priori as known or suspected
cardiovascular risk factors for CHD. Models progressed from less to more rich in covariates:
Model 1: covariates include age, gender, race-ethnicity
Model 2: Model 1 + total cholesterol, HDL cholesterol, smoking status, hypertension, lipidlowering medication
Model 3: Model 2 + education, income, waist circumference, body surface area, BMI, BMI2,
diabetes, LDL, triglycerides
Model 4: Model 2 + metropolitan area
Race/ethnicity was categorized as white non-Hispanic of European ancestry, Chinese, African
American, and Hispanic. BMI was include d as a continuous variable. Cigarette smoking status
was categorized as never, former or current smoker. Annual family income was categorized into
5 categories. Education was classified as: high school not completed, high school completed,
some college but no degree, or completed bachelor’s degree or more. Current use of lipidlowering medications was classified as either some or none. Diabetes was categorized as not
diabetic, impaired fasting glucose, untreated diabetes and treated diabetes. Diabetes was defined
as fasting glucose of ≥ 7.0mmol/L (≥ 126 mg/dL) or use of hypoglycemic medication. Impaired
fasting glucose was defined as fasting glucose =5.5-6.9mmol/L (100-125mg/dL). Hypertension
was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥ 90 mmHg or
taking antihypertensive medications.
Models using the nearest monitor concentration estimates and the base set of covariates
model (Model 2) were considered as the primary models. Sensitivity analysis consisted in
comparing findings across the four models and the three alternative exposure measures, as well
as assessing estimates from selected two-pollutant models and models that controlled for
ultrasound sonographer.
Mean age of the study participants was 62 years, 47.5 percent were male, and 39.1 percent
were non-Hispanic white, 11.7 percent Chinese, 27.4 percent African American, and 21.7
percent Hispanic. Additional characteristics of the study sample are shown in Appendix Table
G.3. The prevalence of current smoking was low (12.7%), and approximately half of the cohort
reported never having smoked.
Median CIMT was 0.84mm (IQR 0.23mm). 49.0% of subjects had a CAC Agatston score >
0; among those, the median score was 86.0 Agatston units (IQR 270.5).
CIMT
Appendix Table G.4 shows estimates of effects of PM2.5 and PM2.5 components on CIMT by
the several exposure estimation methods and analysis models. Appendix Figure G.4 shows
estimated effects on CIMT based on our primary exposure approach (nearest monitor) and
Model 2. Increases in predicted PM2.5, OC, EC and sulfur, but not silicon, were associated with
increased CIMT; CIMT increases per IQR concentration increases were 14.7 µm (95% CI
[9.0,20.5]), 35.1µm (26.8,43.3), 9.6µm (3.6,15.7), 22.7µm (15.0,30.4) and 5.2 µm (-9.8,20.1) for
PM2.5, OC, EC, sulfur and silicon, respectively. The size of the effect estimates for OC and sulfur
was higher than that for EC.
In sensitivity analyses, findings were generally consistent across the three exposure
estimation approaches and were largely unchanged when controlling for more covariates in the
extended model (Model 3). In addition to controlling for lipid-lowering medications in our
primary model, we carried out an analysis restricted to those who reported never having been on
statin medications (n= 4,754); findings in this subgroup were essentially identical to those in the
larger group (results not shown). Effects of adding variables for each metropolitan area to the
model (Model 4), effectively removing between-area effects and allowing assessment of only
within-area effects, were also examined. Metropolitan area variables could not be added to
models in which city-wide average was used as the exposure metric. Several findings were
sensitive to control for area. For our primary exposure method and model, none of associations
of PM2.5 and PM2.5 components with CIMT were significant when metropolitan areas were
included as covariates in model 2 (Model 4) (Appendix Table G.4), although the size of the
effect estimates for EC and sulfur remained essentially unchanged, and the effect of PM2.5 was
only moderately reduced. We also included ultrasound sonographer as an indicator variable in
place of metropolitan area in the CIMT models for sonographers who performed at least 10
studies. Since sonographers were unique to study site, this effectively also controlled for study
area. Results with sonographer in the CIMT models were essentially no different from those
controlling for metropolitan area (results not shown).
For CIMT, estimates from two-pollutant models were examined using nearest monitor and
primary analysis model that included each pair of the PM2.5 components. Only the association of
CIMT with OC was not sensitive to inclusion in the model of the other components or total
PM2.5 (results not shown).
CAC
Appendix Table G.5 shows estimated effects of PM2.5 and PM2.5 components on presence of
CAC by the several estimation methods and models. Appendix Figure G.4 shows estimated
effects on presence of CAC based on our primary exposure approach (nearest monitor) and
Model 2. For this model, there were no statistically significant associations between presence of
CAC and PM2.5 or PM2.5 components. In sensitivity analyses, using IDW or city-wide average,
presence of CAC was negatively associated with EC in Model 2 and in Model 3 with the
extended set of covariates. With adjustment for metropolitan region (Model 4), EC was no longer
negatively associated with presence of CAC.
Appendix Table G.6 shows estimated effects of PM2.5 and PM2.5 components on logtransformed CAC (in those with detectable calcium) by estimation method and models.
Appendix Figure G.4 shows estimated effects on CAC in those with measurable CAC based on
our primary exposure and analysis model. For the primary exposure and analysis model, no
significant positive association of any PM measure and amount of CAC was observed. In
sensitivity analyses, silicon was associated with amount of CAC using city-wide average and
IDW, but in the negative direction; this negative association was no longer present after
adjustment for city region (Model 4).
Appendix Table G.1, Part 1. Participant characteristics at the baseline examination 2000-2002.
Gender
Male
Female
N
%
N
%
2974
3282
Diabetes
Normal
47.5
IFG
52.5
4635
855
74.1
13.7
157
2.5
589
20
9.4
0.3
1057
16.9
1135
1776
2269
19
18.1
28.4
36.3
0.3
5238
1015
3
83.7
16.2
3504
2752
56.0
44.0
Chicago
999
1021
975
982
1088
16.0
16.3
15.6
15.7
17.4
Los Angeles
1191
19.0
Age(years)
45-54
55-64
65-74
75-84
Race-ethnicity
White
Chinese
Black
Hispanic
Income($/year)
<12,000
12,000-24,999
25,000-49,999
50,000-74,999
≥75,000
Missing
Cigarette smoking
Never
Former
Current
Missing
BMI
18.5-22.9
23-27.5
Treated diabetes
1828
1755
1838
835
2449
735
1714
1358
655
1161
1748
1048
1410
234
3145
2299
794
18
29.2
28.1
Untreated diabetes
Missing
29.4 Education
Incomplete high school
13.3
Complete high school
Some college
39.1
Complete college
11.7
Missing
27.4
21.7 Lipid lowering medication
No
Yes
10.5
Missing
18.6
27.9 Hypertension
No
16.8
Yes
22.5
3.7 MESA city
Winston-Salem
New York
50.3
Baltimore
36.7
St. Paul
12.7
0.3
1796
28.7
27.6-40
2459
1777
39.3
28.4
>40
224
3.6 Total
BMI=body mass index; IFG=impaired fasting glucose
6256
0.05
Appendix Table G.1, Part 2. Summary statistics of log(CAC), presence of CAC, and CIMT by gender and race-ethnicity for MESA
exam 1 participants used in the cross-sectional analysis.
Variable
Gender
Race-ethnicity
Total
Log(CAC)
Female
Male
White
Chinese American
Black, African-American
Hispanic
N
1103
1581
1215
325
601
543
2684
Mean
3.96
4.57
4.53
4.03
4.12
4.22
Presence of CAC
SD
1.76
1.83
1.88
1.70
1.81
1.77
N
2872
2621
2179
673
1452
1189
5493
%
38
60
56
48
41
46
N
2872
2621
2179
673
1452
1189
5493
CIMT
Mean
(mm)
0.67
0.69
0.67
0.65
0.72
0.66
SD
0.18
0.20
0.18
0.19
0.20
0.18
Appendix Table G.2. PM2.5 and PM2.5 component annual average concentrations (µg /m3) by city and MESA Air monitor, May 2, 2007 – Apr 16,
2008.
City
LA
monitor ID
L001
L002
LC001
LC002
LC003
Type
non-roadside
Roadside
non-roadside
non-roadside
Roadside
PM2.5
16.33
16.75
13.43
15.31
13.25
EC
1.97
2.17
1.46
1.60
1.40
OC
3.99
3.77
3.02
3.54
2.79
Silicon
0.15
0.16
0.13
0.14
0.12
Sulfur
1.18
1.22
1.20
1.23
1.18
Chicago
C001
C002
C004
non-roadside
non-roadside
non-roadside
12.18
13.66
14.61
1.15
1.27
1.57
3.24
3.15
3.70
0.10
0.13
0.10
1.13
1.19
1.31
C006
C007
non-roadside
Roadside
13.83
15.45
1.31
1.69
3.30
3.71
0.12
0.12
1.26
1.30
B001
B003
B004
Roadside
non-roadside
non-roadside
15.62
14.71
13.86
2.13
1.42
1.36
3.61
3.65
3.25
0.11
0.09
0.08
1.69
1.69
1.67
B005
non-roadside
12.67
0.96
2.90
0.07
1.53
St.Paul
S001
S002
S003
Roadside
non-roadside
non-roadside
10.96
10.23
10.54
1.07
0.68
0.84
3.20
2.67
3.15
0.12
0.11
0.11
0.87
0.85
0.83
NY
N001
N002
non-roadside
Roadside
13.24
15.35
2.32
3.00
3.77
3.96
0.11
0.15
1.46
1.36
Winston-Salem
W001
W002
W003
W004
non-roadside
non-roadside
Roadside
non-roadside
13.22
13.23
13.92
13.01
1.46
1.05
1.22
0.99
4.99
3.51
3.84
3.62
0.09
0.10
0.10
0.09
1.62
1.59
1.66
1.67
Baltimore
*PM2.5, particulate matter<2.5 µm in aerodynamic diameter; EC, element carbon; OC, organic carbon
Appendix Table G.3. Distribution (median and inter-quartile range [IQR]) of PM2.5 and PM2.5 component concentrations (µg/m3) by
three estimation approaches
PM2.5
EC
OC
Silicon
median IQR
median
IQR
median
IQR
median
IQR
Nearest monitor
13.66
2.340 1.36
0.825 3.61
0.724 0.11
0.074
Inverse-distance weighting (IDW) 13.69
1.314 1.32
0.570 3.50
0.624 0.12
0.039
City-wide average
13.57
1.143 1.32
0.509 3.42
0.517 0.11
0.031
*PM2.5, particulate matter<2.5 µm in aerodynamic diameter; EC, element carbon; OC, organic carbon.
Approach
Sulfur
median
IQR
1.30
0.409
1.26
0.425
1.24
0.435
Appendix Table G.4. Difference in CIMT* (µm) for pollutant IQR increases by analysis model and exposure estimation approach
PM2.5*
CIMT difference
(95% CI)
EC*
CIMT difference
(95% CI)
OC*
CIMT difference
(95% CI)
Silicon
CIMT difference
(95% CI)
Sulfur
CIMT difference
(95% CI)
Model 1**
Nearest Monitor
IDW
City-Wide Average
13.7 (8.0,19.5)
8.8 (4.8,12.7)
7.6 (3.7,11.3)
8.2 (2.2,14.2)
4.8 (0.0,9.5)
2.6 (-1.9,7.1)
36.5 (28.3,44.7)
25.4 (19.9,30.8)
21.3 (16.4,26.2)
3.1 (-11.9,18.0)
-0.6 (-10.2,9.0)
-0.2 (-9.9,9.5)
22.6 (14.9,30.2)
23.8 (15.8,31.8)
22.9 (14.8,31.0)
Model 2**
Nearest Monitor
IDW
City-Wide Average
14.7 (9.0,20.5)
9.6 (5.7,13.5)
8.5 (4.7,12.3)
9.6 (3.6,15.7)
6.0 (1.3,10.8)
3.9 (-0.5,8.4)
35.1 (26.8,43.3)
24.9 (19.4,30.3)
20.6 (15.7,25.5)
5.2 (-9.8,20.1)
1.3 (-8.3,10.9)
1.8 (-7.9,11.4)
22.7 (15.0,30.4)
24.0 (16.0,32.0)
23.3 (15.2,31.4)
Model 3**
Nearest Monitor
IDW
City-Wide Average
15.8 (9.9,21.7)
10.6 (6.5,14.6)
9.6 (5.7,13.5)
10.8 (4.6,16.9)
7.2 (2.3,12.0)
5.0 (0.4,9.6)
36.5 (28.0,44.9)
26.9 (21.2,32.6)
21.8 (16.7,26.9)
7.3 (-8.0,22.5)
2.4 (-7.4,12.2)
2.8 (-7.1,12.7)
23.9 (16.0,31.7)
25.4 (17.2,33.5)
24.8 (16.5,33.1)
Model 4**
Nearest Monitor
5.9 (-10.3,22.0)
6.3 (-11.4,23.8)
-2.3 (-26.2,21.5)
-10.1 (-41.0,20.6)
27.4 (-19.3,73.8)
IDW
6.0 (-9.8,21.8)
12.1 (-11.0,35.0)
***
-8.5 (-47.2,30.0)
***
*PM2.5, particulate matter less than or equal to 2.5 µm in aerodynamic diameter; EC, element carbon; OC, organic carbon; CIMT, carotid intimamedia thickness
**Model 1: covariates include age, gender, race-ethnicity
Model 2: Model 1 + total cholesterol, HDL cholesterol, smoking status, hypertension, lipid-lowering medication
Model 3: Model 2 + education, income, waist circumference, body surface area, BMI, BMI2, diabetes, LDL, triglycerides
Model 4: Model 2 + metropolitan area
***unstable estimate
Appendix Table G.5.
CAC* relative risk (RR) for pollutant IQR increases by analysis model and exposure estimation approach.
PM2.5*
EC*
OC*
Silicon
Sulfur
RR (95% CI)
RR (95% CI)
RR (95% CI)
RR (95% CI)
RR (95% CI)
Model 1**
Nearest Monitor
IDW
City-Wide Average
1.00 (0.99,1.01)
0.99 (0.98,1.01)
0.99 (0.98,1.00)
0.96 (0.93,0.99)
0.95 (0.91,0.99)
0.94 (0.90,0.98)
1.03 (0.96,1.09)
0.98 (0.90,1.06)
1.01 (0.09,1.55)
0.84 (0.33,2.14)
0.49 (0.16,1.53)
0.37 (0.09,1.55)
0.98 (0.91,1.07)
0.98 (0.90,1.06)
0.97 (0.89,1.05)
Model 2**
Nearest Monitor
IDW
City-Wide Average
1.00 (0.99,1.01)
1.00 (0.98,1.01)
0.99 (0.98,1.01)
0.97 (0.94,1.00)
0.96 (0.92,1.00)
0.95 (0.91,0.99)
1.01 (0.95,1.07)
0.96 (0.89,1.04)
0.99 (0.91,1.08)
0.92 (0.37,2.32)
0.64 (0.21,1.96)
0.51 (0.12,2.09)
0.98 (0.90,1.06)
0.97 (0.90,1.05)
0.97 (0.89,1.05)
Model 3**
Nearest Monitor
IDW
City-Wide Average
1.00 (0.99,1.01)
1.00 (0.99,1.01)
1.00 (0.98,1.01)
0.97 (0.94,1.01)
0.96 (0.92,1.00)
0.96 (0.91,0.99)
1.03 (0.97,1.09)
0.99 (0.91,1.07)
1.01 (0.93,1.10)
1.02 (0.39,2.64)
0.70 (0.22,2.22)
0.57 (0.13,2.44)
0.99 (0.91,1.07)
0.98 (0.90,1.06)
0.97 (0.90,1.06)
Model 4**
Nearest Monitor
1.01 (0.98,1.05)
1.04 (0.94,1.15)
1.08 (0.91,1.28)
***
1.35 (0.79,2.30)
IDW
1.02 (0.97,1.09)
1.11 (0.91,1.34)
***
***
2.04 (0.62,6.66)
*PM2.5, particulate matter less than or equal to 2.5 µm in aerodynamic diameter; EC, element carbon; OC, organic carbon; CAC, coronary artery
calcification.
**Model 1: covariates include age, gender, race-ethnicity
Model 2: Model 1 + total cholesterol, HDL cholesterol, smoking status, hypertension, lipid-lowering medication
Model 3: Model 2 + education, income, waist circumference, body surface area, BMI, BMI2, diabetes, LDL, triglycerides
Model 4: Model 2 + metropolitan area
***unstable estimate
Appendix Table G.6.
Percentage change in CAC* for pollutant IQR increases by analysis model and exposure estimation approach.
PM2.5*
EC*
OC*
Silicon
% change (95% CI) % change (95% CI)
% change (95% CI)
% change (95% CI)
Model 1**
Nearest Monitor
-1.52 (-3.31,0.26)
-1.65 (-3.54,0.24)
-0.59 (-3.14,1.96)
-3.96 (-8.62,0.70)
IDW
-1.00 (-2.21,0.21)
-1.26 (-2.76,0.24)
-0.61 (-2.29,1.08)
-3.14 (-6.09,-0.19)
City-Wide Average
-0.85 (-2.02,0.32)
-1.11 (-2.52,0.30)
0.19 (-1.32,1.70)
-3.41 (-6.36,-0.45)
Sulfur
% change (95% CI)
0.61 (-1.67,2.89)
0.70 (-1.68,3.08)
0.66 (-1.76,3.07)
Model 2**
Nearest Monitor
IDW
City-Wide Average
-1.56 (-3.34,0.21)
-1.03 (-2.23,0.17)
-0.63 (-1.84,0.58)
-1.61 (-3.49,0.27)
-1.22 (-2.71,0.28)
-0.83 (-2.28,0.63)
-0.98 (-3.52,1.57)
-0.86 (-2.54,0.83)
0.13 (-1.44,1.70)
-3.08 (-7.73,1.58)
-2.52 (-5.47,0.43)
-3.66 (-6.71,-0.61)
0.05 (-2.22,2.33)
0.12 (-2.25,2.50)
1.13 (-1.34,3.60)
Model 3**
Nearest Monitor
IDW
City-Wide Average
-1.37 (-3.22,0.48)
-0.89 (-2.14,0.37)
-0.72 (-1.94,0.49)
-1.43 (-3.37,0.52)
-1.06 (-2.61,0.49)
-0.90 (-2.36,0.57)
-0.86 (-3.51,1.78)
-0.90 (-2.67,0.86)
-0.05 (-1.63,1.53)
-3.35 (-8.18,1.47)
-3.07 (-6.13,0.00)
-3.21 (-6.27,-0.14)
0.58 (-1.76,2.92)
0.67 (-1.77,3.12
0.64 (-1.85,3.12)
Model 4**
Nearest Monitor
-3.10 (-8.08,1.89)
-3.40 (-8.76,1.97)
-1.05 (-8.03,5.93)
4.13 (-5.43,13.68)
-2.32 (-16.91,12.27)
IDW
-3.78 (-8.75,1.19)
-5.42 (-12.52,1.69)
***
3.24 (-8.59,15.07)
8.51 (-24.75,41.77)
*PM2.5, particulate matter less than or equal to 2.5 µm in aerodynamic diameter; EC, element carbon; OC, organic carbon; CAC, coronary artery
calcification.
**Model 1: covariates include age, gender, race-ethnicity
Model 2: Model 1 + total cholesterol, HDL cholesterol, smoking status, hypertension, lipid-lowering medication
Model 3: Model 2 + education, income, waist circumference, body surface area, BMI, BMI2, diabetes, LDL, triglycerides
Model 4: Model 2 + metropolitan area
***unstable estimate
Appendix Figure G.1. PM2.5 and PM2.5 component concentrations (mean and standard deviation bar) by city and exposure estimation
approach. PM2.5, particulate matter <2.5 µm in aerodynamic diameter; EC, elemental carbon; OC, organic carbon; IDW = inverse
distance weighting; Si = silicon; S = sulfur
IDW
0.18
nearest monitor
0.16
city-wide average
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
WinstonSalem
New York
Baltimore
St. Paul
Chicago
city
Silicon concentration by city
IDW
1.8
nearest monitor
1.6
city-wide average
1.4
1.2
1
0.8
0.6
0.4
0.2
0
WinstonSalem
New York
Baltimore
St. Paul
Chicago
city
Sulfur concentration by city
Los Angeles
all
Los Angeles
all
Appendix Figure G.2. Correspondence of mean PM2.5 and PM2.5 component concentrations in
2002 and 2007 from CSN monitoring sites in the MESA cities (https://hei.aer.com/login.php).
PM2.5 = particulate matter <2.5 µm in aerodynamic diameter; EC = elemental carbon; OC =
organic carbon.
Appendix Figure G.3. Estimated effects of PM2.5 and PM2.5 components (per IQR) on difference
in CIMT (µm), relative risk (RR) of CAC, and percent difference in CAC based on nearest
monitor exposure estimates and the base model (Model 2, see Methods). PM2.5 = particulate
matter <2.5 µm in aerodynamic diameter; EC = elemental carbon; OC = organic carbon.
By proximity to MESA monitoring site
Appendix Figure G.4. Cross-sectional effects on CIMT in MESA at exam 1 for an interquartile
increase (0.51, 0.02, 0.89, and 0.69 for sulfur, silicon, EC, and OC, respectively) in predicted
PM2.5 component concentrations from the NPACT spatio-temporal spatial model in six crosssectional models for three prediction areas; 1) participant addresses within 10 kilometers of any
monitor from one year prior to exam 1 through exam 3 participants (primary approach); 2)
address within 5 kilometers; 3) address within 2 kilometers
Appendix Figure G.5: SO4 10-fold CV results. Fitted variogram and predicted vs. observed plots
correspond to using PLS with 2 components as the mean model.
RMSEP = root mean squared error of prediction
PLS=partial least squares
UK= universal kriging; UK Pars= UK parameters
CV= cross-validation
Appendix Figure G.6: SO2 10-fold CV results. Fitted variogram and predicted vs. observed plots
correspond to using PLS with 2 components as the mean model.
RMSEP = root mean squared error of prediction
PLS=partial least squares
UK= universal kriging; UK Pars= UK parameters
CV= cross-validation
Appendix Figure G.7: NO3 10-fold CV results. Fitted variogram and predicted vs. observed plots
correspond to using PLS with 2 components as the mean model.
RMSEP = root mean squared error of prediction
PLS=partial least squares
UK= universal kriging; UK Pars= UK parameters
CV= cross-validation
Appendix Table G.7. Cross-validation statistics for predicted sulfate, nitrate, and SO2 concentrations from
the national model
Nation-wide
MESA Air 6 city area*
PLS + Universal
kriging
PLS only
PLS + Universal
kriging
PLS only
RMSE
R2
RMSE
R2
RMSE
R2
RMSE
R2
Sulfate
0.22
0.47
0.08
0.93
0.17
0.00
0.06
0.86
Nitrate
0.34
0.06
0.17
0.76
0.34
0.00
0.15
0.79
SO2
0.41
0.31
0.36
0.45
0.34
0.18
0.30
0.35
* Defined by 200 kilometer buffers from each city center
# Cross-validation statistics for NO2 predictions from the spatio-temporal model by city is shown in
Appendix C
Cross-sectional and longitudinal analyses for sulfate, nitrate, SO2, and NO2
Appendix Figure G.8. Predicted long-term concentrations of sulfate, nitrate, and SO2 from the national
spatial model, and NO2 from the spatio-temporal model at participant locations by 6 cities (different
colors represent quintiles of the range of concentrations for a component in each city; blue, green, yellow,
orange, and red display 1st, 2nd, 3rd, 4th, and 5th quintiles).
Appendix Figure G.9. Cross-sectional associations for presence of CAC, log(CAC) and CIMT in MESA
at exam 1 for an interquartile increase (0.56, 1.65, 1.14, and 13.46 for sulfate, nitrate, SO2, and NO2,
respectively) in predicted sulfate, nitrate, and SO2 concentrations from the national spatial model and NO2
from spatio-temporal model in six cross-sectional models. (See Table 38 in the Section 1 main text for
description of the six models.)
Appendix Figure G.10. Cross-sectional and longitudinal associations from the longitudinal model for
CIMT in MESA for an interquartile increase (0.56, 1.65, 1.14, and 13.46 for sulfate, nitrate, SO2, and
NO2, respectively) in predicted sulfate, nitrate, and SO2 concentrations from the national spatial model
and NO2 from spatio-temporal model in six models. (See Table 38 in the Section 1 main text for
description of the six models.)
Cross-sectional and longitudinal analyses adjusting for NO2 and SO2
Appendix Figure G.11. Cross-sectional associations for presence of CAC, log(CAC) and CIMT
in MESA at exam 1 for an interquartile increase (1.51, 0.51, 0.02, 0.89, and 0.69 for PM2.5,
sulfur, silicon, EC, and OC, respectively) in predicted PM2.5 and PM2.5 component concentrations
from the NPACT spatio-temporal model in the primary model (model 3) adjusting for NO2
(model 7) and SO2 (model 8)
Appendix figure G.12. Cross-sectional and longitudinal associations from the longitudinal model
for CIMT in MESA for an interquartile increase (1.51, 0.51, 0.02, 0.89, and 0.69 for PM2.5,
sulfur, silicon, EC, and OC, respectively) in predicted PM2.5 and PM2.5 component concentrations
from the NPACT spatio-temporal model in the primary model (model 3) adjusting for NO2
(model 7) and SO2 (model 8)
Appendix Figure G.13. Cross-sectional associations for presence of CAC, log(CAC) and CIMT
in MESA at exam 1 for an interquartile increase (2.19, 0.18, 0.02, 0.28, and 0.39 for PM2.5,
sulfur, silicon, EC, and OC, respectively) in predicted PM2.5 and PM2.5 component concentrations
from the national spatial model in the primary model (model 3) adjusting for NO2 (model 7) and
SO2 (model 8)
Appendix figure G.14. Cross-sectional and longitudinal associations from the longitudinal model
for CIMT in MESA for an interquartile increase (2.19, 0.18, 0.02, 0.28, and 0.39 for PM2.5,
sulfur, silicon, EC, and OC, respectively) in predicted PM2.5 and PM2.5 component concentrations
from the national spatial model in the primary model (model 3) adjusting for NO2 (model 7) and
SO2 (model 8)
Cross-sectional effects of sulfur on CAC and CIMT with the interaction of traffic variables
Appendix Table G.8. Cross-sectional association for presence of CAC, log(CAC) and CIMT in
MESA at exam 1 for an interquartile increase in predicted of sulfur from spatio-temporal model
and national spatial with the interaction with NO2 and road proximity variable
Presence of CAC
Log CAC
CIMT
Exposure model
Health model
Pollutant RR* 95% CI exp(B) 95% CI
B
95% CI
Spatio-temporal Interaction with NO2 Sulfur 1.028 0.943 1.121 0.892 0.689 1.156 0.016 -0.002 0.034
S*NO2 0.996 0.990 1.003 1.012 0.994 1.031 -0.002 -0.003 -0.001
Interaction with road+ Sulfur 0.985 0.944 1.028 1.029 0.904 1.171 -0.006 -0.015 0.003
S*road 0.976 0.892 1.068 0.995 0.772 1.283 -0.003 -0.020 0.015
National Spatial Interaction with NO2 Sulfur 1.002 0.926 1.085 0.890 0.705 1.123 0.011 -0.005 0.027
S*NO2 1.000 0.994 1.005 1.013 0.997 1.030 -0.001 -0.003 0.000
Interaction with road Sulfur 0.994 0.964 1.025 1.049 0.957 1.150 -0.007 -0.013 0.000
S*road 1.016 0.951 1.085 1.030 0.859 1.236 0.001 -0.012 0.014
* Effect estimates (RR and beta coefficients) and 95% CIs were presented per interquartile
increase in sulfur: 1.51 for the spatio-temporal model and 0.18 for national spatial model
+ Indicator variable for proximity to roads defined by addresses within 100 meters from the
closest A1 or A2 road or 50 meters from the closest A3 road